Chisle Audit Agent Skill

Chisle Audit

Scan the target (a diff, a file, or the repo tree) and report what to cut, on

both axes. One-shot. Read-only: never edit, never write a flag, never apply fixes.

Scope

• No argument → audit the current git diff (staged + unstaged). Empty diff → audit HEAD~1..HEAD.
• A path → audit that file or directory.
• "repo" / "whole repo" → walk the tree (skip vendored/generated/node_modules/dist/lockfiles).

What to flag

Code (the YAGNI axis):

• Reinvented stdlib (hand-rolled debounce, deep-clone, groupBy, retry loop, date math)
• Abstraction with one implementation (interface/factory/wrapper for a single case)
• New dependency for what a few lines or an installed dep already covers
• Config/option/flag that never varies
• Speculative "for later" scaffolding with no current caller
• Verbose code where a native platform feature (CSS, DB constraint, <input type>) does it

Prose (the compression axis), the half a code-only auditor misses:

• Comments that restate the code (i += 1 // increment i)
• Docstrings/READMEs padded with filler, hedging, ceremony, or duplicated content
• Multi-paragraph explanations where one tight sentence carries the meaning
• Decorative tables/emoji/headings that add tokens, not information
• Dead prose: TODO graveyards, stale "see also" links, obsolete sections

What NOT to flag

Input validation at trust boundaries, error handling that prevents data loss,

security, accessibility, deliberate // chisle: / // ponytail: shortcuts already

documented, or domain comments that explain *why* (not *what*).

Output

One ranked list, biggest cut first. One finding per line. No preamble, no praise.

path:line  [code|prose]  <what's bloated> → <the lean replacement>. (~N lines/tokens)

End with a two-line summary:

N findings: X code, Y prose. Est. removable: ~A lines code, ~B lines prose.
Biggest win: <the single highest-impact cut>.

Be honest about uncertainty: mark a finding (check) if cutting it might lose

behavior you can't verify from the snippet. Lean toward fewer, high-confidence

findings over a long speculative list.